Incidence and risk factors for extended post-operative length of stay following primary hip arthroplasty in a South African setting

Background: This study sought to determine the incidence of extended post-operative length of stay (EPLoS) and its associated risk factors in South African primary hip arthroplasty patients. Methods: This was a retrospective chart review study of 185 adults who underwent primary hip arthroplasty at a quaternary South African hospital. Data related to patient, clinical, and surgical characteristics were collected. Post-operative length of stay was calculated as the time (in days) between the dates of surgery and discharge from hospital. We defined EPLoS as any length of stay ≥75th percentile obtained for the entire study population. Data were analysed using univariate and multivariate statistical methods. Results: The incidence of EPLoS was 28.1% (95% confidence interval – CI: 22.1–35.0%). Risk factors for EPLoS included: female sex (odds ratio – OR: 4.63, 95% CI: 1.74–12.34; p=0.002), patient’s maximum walking distance <100 m (OR: 3.05, 95% CI: 1.05–8.89; p=0.041) and extended duration of surgery (OR: 3.62, 95% CI: 1.31–10.01; p=0.013). Conclusion: We provide a report of EPLoS and several associated risk factors in South African primary hip arthroplasty patients.


Introduction
Increased global life expectancy has been linked to a higher burden of musculoskeletal conditions, including hip fracture and osteoarthritis. 1Untreated musculoskeletal conditions impact quality of life in afflicted patients and also have adverse consequences on healthcare expenditure and resource utilisation. 1 These conditions would therefore have public health significance in resource-limited settings.Aside from non-communicable aetiologies, the global HIV epidemic has also been linked to the growing prevalence of orthopaedic disorders. 2Conservative medical therapy might not be effective in a large proportion of patients afflicted with orthopaedic hip conditions.Surgical intervention remains the only viable management option in these patients. 3The effectiveness of primary hip arthroplasty in reversing pain and loss of function associated with orthopaedic hip conditions is well described. 3,4Utilisation of primary hip arthroplasty as a surgical intervention for orthopaedic hip conditions has increased substantially over the past two to three decades, with this procedure now considered among the most common surgical procedures performed worldwide. 4 survey of orthopaedic surgeon members belonging to the South African Orthopaedic Association reported that each member in the country performed up to 43 hip arthroplasties each year.5 In addition, a lack of surgical expertise and other essential resources in surrounding countries has resulted in a number of patients from these countries being referred to South African hospitals for the procedure.6 In response to the increasing demand for primary hip arthroplasty, it is possible that many South African orthopaedic surgery units will in future adopt accelerated post-operative care pathways, in which the length of inpatient stay (and subsequent expenditure and resource utilisation for each patient) following surgery is reduced.7,8 An understanding of which patient, clinical, and surgical characteristics are associated with extended postoperative length of stay (EPLoS) in South African primary hip arthroplasty patients would have important future implications for the development of fast-track or accelerated surgical and recovery protocols implemented at orthopaedic surgery units in the country.
Therefore, the objectives of this study were to: 1. Determine the incidence of EPLoS in a sample of South African primary hip arthroplasty patients 2. Determine which patient, clinical, and surgical characteristics are associated with EPLoS in a sample of South African primary hip arthroplasty patients.

Study design, study setting, and study population
This was a retrospective chart review study involving consecutive adult patients who were admitted for primary hip arthroplasty through a dedicated arthroplasty unit at a quaternary level hospital in KwaZulu-Natal, South Africa, between 23 September 2014 and 28 July 2016.Inclusion/exclusion criteria for this study are presented in Table I. Potential participants were identified from theatre lists during the specified study period.

Data collection
The medical records of all patients included in this study were reviewed and data related to various patient (demographics), clinical (comorbidities, presenting diagnosis, Thomas test with fixed flexion deformity [FFD], etc.), and surgical characteristics (nature of surgery, anaesthesia, surgical approach, duration of surgery, and peri-operative blood transfusion) were collected using case report forms.We also collected data related to the occurrence of serious peri-operative complications, which we defined as a grade III or above peri-operative complication when using the Clavien-Dindo classification (includes: organ failure, critical care admission, re-operation, and mortality). 9Post-operative length of stay was calculated as the time (days) between the date of a patient's operation and the date that the patient was discharged from hospital.
The study outcome was EPLoS.This was defined as a postoperative length of stay ≥75th percentile calculated for the entire study population.This definition of EPLoS has been used in similar surgical studies conducted in overseas settings. 10,11Data were transferred from the case report forms to a Microsoft Excel ® spreadsheet in preparation for analysis.

Data analysis
The median length of stay for the study population was calculated and is presented with an interquartile range.The incidence of EPLoS in this study was calculated using conventional epidemiological methods. 12The incidence of EPLoS in this study is presented as a percentage with 95% confidence intervals (95% CI).Potential associations between various patient, clinical, and surgical characteristics and EPLoS were investigated using univariate (χ 2 test, or Fisher's exact test) and multivariate (binary logistic regression) statistical methods.Results for the univariate statistical analysis are presented as frequencies and percentages.Characteristics with p<0.100 in the univariate analysis were selected for inclusion in the multivariate statistical analysis.This purposeful selection of variables for inclusion in the multivariate analysis was done to obtain the most parsimonious model possible. 13Model fit was assessed using a Hosmer-Lemeshow test.Results for the multivariable statistical analysis are presented as odds ratios (OR) with 95% CI.A p-value of <0.050 was considered to be a statistically significant result.All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 24.0 (IBM Corp, USA).

Derivation of study population and incidence of EPLoS
The derivation of the study population and the incidence of EPLoS in this study is shown in Figure 1.Following the application of our study inclusion and exclusion criteria, our final study population consisted of 185 adult patients who underwent primary hip arthroplasty.The median post-operative length of stay for the study population was 5.0 days (interquartile range: 3.0-7.0days).The 75th percentile for the study population post-operative length of

Distribution of patient, clinical, and surgical characteristics in the study population
The distribution of patient, clinical, and surgical characteristics in the study population is shown in Table II The median duration of surgery for the study population was 100.0 minutes (interquartile range: 75.0-125.0minutes).Urgent/ emergent surgical procedures were rare in the study population (n/N=3/185, 1.6%).Surgery was performed under general anaesthesia in 96/185 patients (51.9%).The standard posterior approach was used in 123/185 procedures (66.5%).Thirty-five patients (18.9%) experienced surgery of extended duration, which we defined as a surgery with a duration ≥75th percentile calculated for the entire study population.Twenty-six patients in the study population required a peri-operative blood transfusion (14.1%).

Results of the multivariable statistical analysis
The results of the multivariable statistical analysis are shown in Table III.Only nine of the characteristics investigated in the univariate analysis met the criteria of p<0.100 for inclusion in the multivariable analysis.These characteristics were sex, being a current smoker, hypertension, presenting diagnosis, FFD, patient's maximum walking distance, extended duration of surgery, peri-operative blood transfusion, and serious peri-operative complications.Of these characteristics, only three were found to be independently associated with EPLoS.These characteristics were female sex (when compared with males, OR:

Discussion
1][22][23] This has resulted in a growing demand for hip arthroplasty in the country. 6However there are staffing and economic challenges in running orthopaedic surgical units in the public sector, 24 and the availability of beds in these public hospitals might also be a concern. 25In order to cope with the higher demand for hip arthroplasty, some hospitals are beginning to implement fast-track protocols which are aimed at reducing post-operative length of stay while minimising the rate of post-discharge complications in suitable patients who undergo the surgical procedure. 265][16] With regard to EPLoS following hip arthroplasty, the literature is scant.However, there is one American study which reported EPLoS in this surgical population. 10In that study, the 75th percentile for the population post-operative length of stay was 14.0 days, which is twice that reported for our study. 10Furthermore, one-third of the American study population experienced EPLoS. 10s with our findings for median post-operative length of stay, the discrepancy in EPLoS between the American study population and our South African study population must be viewed in the context of a growing demand for hip arthroplasty in South Africa and the disproportionate availability of healthcare resources between South African and American settings. 27e found statistically significant univariate associations between several characteristics (including: sex, presenting diagnosis, FFD, patient's maximum walking distance, extended duration of surgery, peri-operative blood transfusion, and serious post-operative complications).These findings are not unique to our study.Other overseas studies have reported univariate statistical associations between these/similar characteristics and post-operative length of stay in hip arthroplasty patients. 14,16We found three characteristics to be independently associated with EPLoS (including: sex, patient's maximum walking distance and extended duration of surgery).Female sex was found to be associated with an almost five-fold increase in the risk of experiencing EPLoS following primary hip arthroplasty.Abbas et al., reported an almost two-fold increase in the risk of EPLoS for women undergoing hip arthroplasty in a Pakistani setting. 16Dall et al., also reported a multivariate statistical association (without describing the magnitude of risk) between female sex and longer post-operative length of stay a British hip arthroplasty population. 14Therefore, our findings for female sex appear, in general, to be in agreement with the published literature.However, the difference in the magnitude of odds ratios for female sex obtained in our study and the study of Abbas et al. 16 requires further investigation.The characteristics of patient's maximum walking distance have not been specifically investigated as potential risk factors for EPLoS following hip arthroplasty in the published literature.However, these characteristics are components of the pre-operative Harris Hip Score, 28 which has been shown by Dall et al., 14 to be associated with length of stay following hip arthroplasty.Specifically, these characteristics appear to be associated with mobility and functional status in patients with hip conditions. 28herefore, our findings highlight the potential importance of preoperative functional status and ambulation on the post-operative recovery period in South African primary hip arthroplasty patients.Lastly, we found extended duration of surgery to be associated with an almost four-fold higher risk of experiencing EPLoS.This is somewhat in agreement with the British study of Foote et al., who also report extended duration of surgery to be independently associated with a higher risk of EPLoS. 15However, as with the patients' sex, there appears to be a difference in the magnitude of odds ratios for surgery duration between our study and the study of Foote and colleagues. 15Attempts to should be made to reduce the duration of hip arthroplasty in our setting, possibly through the application of benchmarks and optimisation of surgical technique.
The risk factors identified in our study can be incorporated into future risk stratification systems for EPLoS in South African orthopaedic units.Similar risk stratification systems based on identified risk factors for EPLoS following primary hip arthroplasty have been proposed by Abbas et al., 16 and Foote et al. 15 These risk stratification systems are required to be developed and validated for performance in a separate surgical cohort. 29This step is beyond the scope of the dataset used in our study and requires further research.
There were several characteristics which were not found to be associated with EPLoS during the univariate statistical analysis, or following inclusion in the multivariable statistical analysis.There are two explanations for the lack of statistical association between these characteristics and EPLoS in our study.First, it might be possible that these characteristics, while identified as risk factors in overseas settings, are genuinely not associated with EPLoS in South African hip arthroplasty patients.Discordance in clinical risk factors between overseas/South African surgical populations and other post-operative outcomes has been described elsewhere. 30t might be worthwhile to involve overseas collaborators with access to overseas patient data in future research such that valid comparisons of risk factors between our settings can be made.Secondly, it is possible that a larger sample size than 185 patients would be required to investigate the impact of these characteristics on EPLoS.A potential solution to this would be a collaborative study involving as many hospitals which offer orthopaedic surgical services as possible.
Our study had several strengths.Our study is, to the best of our knowledge, the only South African study which specifically investigates EPLoS following primary hip arthroplasty.Another strength of our study is that we included data on HIV infection in our statistical analyses.This is important as the prevalence of HIV is usually much lower in American and British populations, 19 and so our study provides important information on the impact of this characteristic in settings with a high burden of HIV infection.The final strength of our study is that while our sample size appeared modest, it still allowed for us to perform a multivariable statistical analysis to determine independent risk factors for EPLoS without any serious violation of statistical rules of thumb. 31Our study also had several limitations.First, as this study was conducted at a single, dedicated arthroplasty unit in a quaternary level hospital with standardised pre-and post-operative protocols in place, it might be argued that our study findings lack generalisability.As for our solution for the challenge related to the lack of statistical association between several characteristics and EPLoS, we recommend that collaborative studies involving hospitals at various levels of service delivery are conducted to determine the generalisability of our study findings.In addition, we were unable to investigate the impact of the Harris Hip Score in our study due to poor documentation of this characteristic in the patient medical records.We did, however, find that components of the Harris Hip Score were statistically associated with EPLoS, and it is therefore possible that the composite Harris Hip Score might also be associated with EPLoS.Prospective research wherein data collection for the Harris Hip Score is standardised is required.Finally, we did not report the impact of EPLoS on healthcare expenditure or post-discharge complications.These outcomes can only be appropriately investigated through the conduct of prospective research studies.
In conclusion, we found several risk factors for EPLoS following primary hip arthroplasty in South African patients.These risk factors included sex, patient's maximum walking distance, and extended duration of surgery.Further research is required to confirm our study findings, as well as address the limitations identified in our study.

Table I :
Inclusion and exclusion criteria for the proposed study stay was 7.0 days.A total of 52/185 patients experienced EPLoS following primary hip arthroplasty, with the calculated incidence of EPLoS being 28.1% (95% CI: 22.1-35.0%).

Table II :
Distribution of patient/clinical characteristics in the study population and results of the univariate statistical analysis* *Results expressed as frequencies (%).p<0.050 was considered a statistically significant result EPLoS: extended post-operative length of stay; ASA: American Society of Anesthesiologists; COPD: chronic obstructive pulmonary disease; UC: unable to compute; CNBE: could not be established; FFD: fixed flexion deformity; VAS: visual analogue score

Table III :
Results of the multivariate statistical analysis* *Results adjusted for confounders.Only characteristics with p<0.100 in the univariate statistical analysis included in the multivariable statistical analysis.p<0.050 was considered a statistically significant result.OR: odds ratio; CI: confidence interval; FFD: fixed flexion deformity; CNBE: could not be established Table II continued from page 43